106 
Fishery Bulletin 113(2) 
Brodeur, 2006; Brodeur et al., 2008) to estimate sur- 
vey and fishery catch outside the sampled domain. 
To assess ordinary kriging model performance, model 
diagnostics (including root mean square error of pre- 
diction [RMSE] and variance estimates from 100-fold 
cross-validation) were investigated (Cressie, 1993). All 
analyses were carried out in R with the gstat pack- 
age (Pebesma, 2004). After interpolation of survey and 
fishery catch throughout the NES LME, geostatistical 
spatial overlap of predicted spiny dogfish distribution 
with predicted fishery catch (SOj) was calculated with 
the following equation: 
SOA%) = ^^-xlOO, (5) 
N Fl 
where iVsi,FI = the number of grid cells with predicted 
survey and fishery catch of spiny dog- 
fish; and 
Np i = the number of grid cells with predicted 
fishery catch. 
As with SOq, this metric provides insight regarding 
the area where both the commercial fishery and bot- 
tom trawl survey were predicted to catch spiny dog- 
fish but does so over a larger spatial scale. Here, low 
overlap indicated that the fishery caught spiny dogfish 
that were not accounted for by the bottom trawl survey. 
The footprint of the SO\ encompassed predicted fishing 
grounds inhabited by spiny dogfish. As explained previ- 
ously for SOq and SOp, sites that were surveyed but 
not fished had no bearing on SOp 
Availability to fishery 
The percentage of spiny dogfish stock available to each 
commercial fishery was used to infer changes in catch- 
ability of the population, in the sense that increased 
availability could lead to increased catchability. Annual 
estimates were obtained with the following equation: 
where CF = the total survey catch of spiny dogfish 
in grid cells where commercial fishing oc- 
curred; and 
CT - the total survey catch of spiny dogfish. 
Availability was reported both overall (i.e., for all spiny 
dogfish) and separately for each life-history stage be- 
cause of sex-specific trends in habitat preference and 
distribution (Sagarese et al., 2014b). The life-history 
stages that were examined included aggregated male 
and female neonates (<26 cm in total length [TL] ), im- 
mature males (>26 cm TL and <60 cm TL), immature 
females (>26 cm TL and <80 cm TL), mature males (>60 
cm TL), and mature females (>80 cm TL). We assumed 
that the survey catch was representative of trends for 
the entire spiny dogfish stock and also for individual 
life-history stages throughout the NES LME. A high 
percentage indicated that a large portion of the spiny 
dogfish stock was present in areas represented by the 
grid cells where commercial fisheries were operating. 
For each fishery, the relationship between availability 
and log e -transformed abundance of spiny dogfish was 
examined with annual values considered observations. 
This analysis focused on mature females because this 
stage is preferentially landed for maximal profit in the 
highly sex-selective fishery for spiny dogfish (Rago et al., 
1998) and also because of their close proximity to shore 
(Sagarese et al., 2014b). Between 1982 and 1995, 95% of 
sampled landings were mature females (NEFSC 3 ). 
Correlation between fisheries and survey catch data 
We investigated annual correlations between com- 
mercial fisheries (effort and catch) and survey catch 
of spiny dogfish (Park and Obrycki, 2004). Map-cor- 
relation coefficients were calculated on the basis of 
sample-to-sample correlations in which grid cells were 
treated as samples. Pearson’s coefficient of correlation 
(r) measured the “strength” of the relationship, and 
Spearman’s rank correlation coefficient (r sp ) measured 
the similarity between ranks of observed and predicted 
values (Quinn and Keough, 2002). 
Results 
Comparison of data from different gears 
Preliminary analysis of observer data from the NE- 
FOP revealed that the majority of spiny dogfish were 
captured by the SGN (autumn: 57%; spring: 47%) and 
OT fisheries (autumn: 29%; spring: 45%). Therefore, all 
analyses focused on these 2 fisheries. Data for these 
fisheries covered the longest and most continuous time 
series (1989-2010). The SGN fishery expended more ef- 
fort (in hours fished) and kept a larger percentage of 
spiny dogfish catch (autumn: 61%; spring: 82%) than 
the OT fishery (autumn: 8%; spring: 6%). Annual dis- 
cards (%) of spiny dogfish for the OT fishery generally 
exceeded 85% during both seasons, although a few ex- 
ceptions were noted in the mid- to late 1990s (range: 
from 20% [1996] to 68% [1999]). In contrast, annual 
discards of spiny dogfish for the SGN fishery were 
highly variable during both seasons (spring: 1-100%; 
autumn: 1-90%) but exhibited consistently low values 
during the mid- to late 2000s. Interannual differences 
existed in the percentage of discarded catch, likely be- 
cause of the complex history of dogfish harvest — which 
began in 1990, was shut down in 1999, and was pri- 
marily bycatch in the 2000s. 
Estimates of mean percent observer coverage (the 
number of trips observed compared with the number 
of commercial trips reported) were as follows: north- 
east SGN (1990-2008: 5.1% [standard deviation (SD) 
1.7]), mid-Atlantic SGN (1995-2008: 2.9% [SD 1.3]), 
northeast bottom trawl (including beam, bottom fish 
[otter], bottom shrimp, and bottom scallop trawls) 
(1994-2008: 3.2% [SD 3.6]), and mid-Atlantic bottom 
